unsloth
Mistral Small 3.2 24B Instruct 2506
Mistral Small 3.2 24B Instruct 2506 is a 23.6B dense llama model from unsloth. Configured context length is 131,072 tokens. License: apache-2.0. Quant sizes are the GGUF file bytes published on Hugging Face when a file exists.
Parameters23.6B
Context128K
Attentiongqa
Licenseapache-2.0
40 layers · 8 KV heads · head dim 128 · GGUF repo · config
Quant file sizes
| Quant | File size | Source |
|---|---|---|
| BF16 | 43.9 GB | Hugging Face file bytes |
| UD-Q8_K_XL | 27.0 GB | Hugging Face file bytes |
| Q8_0 | 23.3 GB | Hugging Face file bytes |
| UD-Q6_K_XL | 19.4 GB | Hugging Face file bytes |
| Q6_K | 18.0 GB | Hugging Face file bytes |
| UD-Q5_K_XL | 15.6 GB | Hugging Face file bytes |
| Q5_K_M | 15.6 GB | Hugging Face file bytes |
| Q5_K_S | 15.2 GB | Hugging Face file bytes |
| UD-Q4_K_XL | 13.5 GB | Hugging Face file bytes |
| Q4_K_M | 13.3 GB | Hugging Face file bytes |
| Q4_K_S | 12.6 GB | Hugging Face file bytes |
| IQ4_NL | 12.5 GB | Hugging Face file bytes |
| IQ4_XS | 11.9 GB | Hugging Face file bytes |
| Q4_1 | 13.9 GB | Hugging Face file bytes |
| Q4_0 | 12.6 GB | Hugging Face file bytes |
| UD-Q3_K_XL | 11.0 GB | Hugging Face file bytes |
| Q3_K_M | 10.7 GB | Hugging Face file bytes |
| Q3_K_S | 9.69 GB | Hugging Face file bytes |
| UD-IQ3_XXS | 8.76 GB | Hugging Face file bytes |
| UD-Q2_K_XL | 8.65 GB | Hugging Face file bytes |
| Q2_K | 16.7 GB | Hugging Face file bytes |
| UD-IQ2_XXS | 6.29 GB | Hugging Face file bytes |
| UD-IQ1_M | 5.60 GB | Hugging Face file bytes |
| UD-IQ1_S | 5.18 GB | Hugging Face file bytes |
On common machines
| Device | Verdict | Quant | Speed |
|---|---|---|---|
| GeForce RTX 4060 Ti 8GB | Tight fit | UD-IQ1_M | 24–32 |
| GeForce RTX 3060 12GB | Tight fit | UD-IQ3_XXS | 18–25 |
| GeForce RTX 4060 Ti 16GB | Tight fit | Q4_K_M | 11–14 |
| GeForce RTX 4090 | Tight fit | UD-Q6_K_XL | 25–34 |
| GeForce RTX 5090 | Tight fit | UD-Q8_K_XL | 32–44 |
| Apple M4 Pro | Tight fit | UD-Q5_K_XL | 8.3–11 |
| Apple M5 | Tight fit | Q3_K_S | 7.6–10 |
| CPU only · DDR5-5600 | Runs on CPU | Q8_0 | 1.9–2.6 |
Smallest devices that fit it fully
- GeForce RTX 5060 Ti 8GB — Tight fit, UD-IQ1_M, 37–50 tok/s
- GeForce RTX 5060 — Tight fit, UD-IQ1_M, 37–50 tok/s
- GeForce RTX 5050 — Tight fit, UD-IQ1_M, 26–36 tok/s
- GeForce RTX 4060 Ti 8GB — Tight fit, UD-IQ1_M, 24–32 tok/s
- GeForce RTX 4060 — Tight fit, UD-IQ1_M, 22–31 tok/s
- GeForce RTX 3070 Ti — Tight fit, UD-IQ1_M, 50–68 tok/s
Run it
Commands for the GeForce RTX 4060 Ti 16GB estimate (Q4_K_M).
llama.cpp
llama-server -hf unsloth/Mistral-Small-3.2-24B-Instruct-2506-GGUF:Q4_K_M -c 4096 -ngl 40
Ollama
ollama run mistral
LM Studio
lms get unsloth/Mistral-Small-3.2-24B-Instruct-2506-GGUF # Then load quant Q4_K_M